[ Home ] [ Meet the Authors ] [ Errata Corrige ] [ Suggestions ] [ Table of Contents ] [ Downloads ] [ VEX forum ]

Embedded Computing

A VLIW Approach to Architecture, Compilers and Tools

Joseph A. Fisher, Paolo Faraboschi, Cliff Young

 

Errata Corrige

PAGE Error Correction / Comment

xxix

wires are expensive, whereas transitors are cheap

wires are expensive, whereas transistors are cheap
xxxi

Memories in Mutliprocessor Systems

Memories in Multiprocessor Systems
1 In 1943, the chairman of IBM predicted a world market for no more than five computers.

Although Watson is well known for his alleged 1943 statement: "I think there is a world market for maybe five computers," there is no evidence he ever made it. See this wikipedia entry on the subject.
9

matrix math extensions (MMX) in Pentium processors

It is not clear what MMX really stands for. One can find MultiMedia eXtension, Multiple Math eXtension, or Matrix Math eXtension. However, it is likely a trademark by Intel with no specific meaning. See this wikepedia entry on the subject.
39 the book by Wolf [2000] the book by Wolf [2001]
40 the book by Ashenden [1995] the book by Ashenden [2001]
40 the books ... by Ball [1998, 2001] the books ... by Ball [1998, 2003]
40 Rebecca Henderson's [Henderson, 2004] commentary Rebecca Henderson's [Henderson, 2000] commentary
61 (fig. 2.4)

FP Funtional Units ... Integer Funtional Units

 FP Functional Units ... Integer Functional Units
75 CDC-7600 [Weis and Smith, 1987; CDC-7600 [Weiss and Smith, 1987;
78 (ex. 5) First suggested by Fisher et al. [Fisher, 1981] First suggested by Fisher et al. [Fisher et al., 1981]
112 There is no "Variable" row in the table ... There is no "Dynamic" row in the table ...
141 (example) each byte of the inputs registers each byte of the input registers
170 (example) the add always executes (speculatively) the mpy always executes (speculatively)
174 Goldberg [1990] Goldberg [1996]
225 Johnson [1992] Johnson [1991]
226 Mahlke et al. [1994] Mahlke et al. [1993]
226 Palacharla, Jouppi, and Smith [1993] Palacharla, Jouppi, and Smith [1997]
226 Farkas, Chow, Jouppi, and Vranesic [1997] Farkas, Chow, Jouppi, and Vranesic [1997a]
226 Zyubian and Kogge [1998] Zyuban and Kogge [1998]
226 Chang et al. [1991] Chang et al. [1991b]
226 Ozer et al. [1998] Ozer et al. [1998a]
226 Ranghunathan, Jha, and Dey [1998] Raghunathan, Jha, and Dey [1998]
282 VSI [2000] VSI [2004]
282 the book by Wolfe [2002] the book by Wolf [2002]
283 the book by Rashihnkar, Paterson and Singh [2000] the book by Rashihnkar, Paterson and Singh [2001]
283 paper by Walls and Williams [2002] paper by Walls and Williams [2000]
283 paper by Chaouni, Cyr, Giacalone et al. paper by Chaoui, Cyr, Giacalone et al .
283 by Conte and Gimare [1995] by Conte and Gimarc [1995]
284 work by Goldsmith and Hennessy [1993] work by Goldschmidt and Hennessy [1993]
284 Maurer and Wang [1990] Maurer and Wang [1991]
284 Reshadi, Misra, and Dutt [2003] Reshadi, Mishra, and Dutt [2003]
331 the VEX compiler can also be called "parametrizable" the VEX compiler can also be called "parameterizable"
332 Aho, Sethi, and Ullmann [1986] Aho, Sethi, and Ullman [1986]
332 Allen and Kennedy [2002] Allen and Kennedy [2001]
332 Liao et al. [1995a and b] Liao et al. [1995 and 1996]
332 Laneer et al [1995] Lanneer et al [1995]
332 Hanono and Devadas [1998] Hadjiyianni, Hanono and Devadas [1997]
333 Rabbah, Palem et al [2003] Rabbah and Palem [2003]
333 Roy and Johnson [1996] Roy and Johnson [1997]
335 (ex.10) support for soft- ware pipelining support for software pipelining
360 block B is control dependent on block A if all paths from the start of the program to B pass through A and some path from A to the exit of the program omits B block B is control dependent on block A if there is a path from A that will always cause B to be executed, and another path from A that may avoid B
360 (fig 8.11) double-headed arrow between "add t4 = t3,1" and "sub t4 = t1,1" single-headed arrow from "add t4 = t3,1" to "sub t4 = t1,1"
366 (fig. 8.14) In 8.14(b), the "Z" top entry on the left portion should be "X" and not "Z", as it is evident from the scheduled code
440 Operating Systems: Internals and Design Principles[Stallings, 2000] Operating Systems: Internals and Design Principles[Stallings, 2004]
440 see Lions' commentary on UNIX, sixth edition [Lyons, 1996] see Lions' commentary on UNIX, sixth edition [Lions, 1996]
456 ISO/ANSO C standard ISO/ANSI C standard
465 #define __ADDCG(sum,cout, a,b,cin) { \ #define __ADDCG__(sum,cout, a,b,cin) { \
465 r = __asm(__ADDCG __,a,b,cin); r = __asm(__ADDCG__,a,b,cin);
474 As J2ME (...) becomes popular, ... , an additional challenge in to evaluate how these devices perform.. As J2ME (...) becomes popular, ... , an additional challenge is to evaluate how these devices perform..
483 Qin and Malek [2003] Qin and Malik [2003]
489 Ellis and Stroustrup [1997] Ellis and Stroustrup [1997]
490 Wolfe, Ortega, and Shanklin [1996] Wolfe [1996]
490 Kennedy and Allen [2002] Allen and Kennedy [2002]
498 the transformation needs to the strongly nonlinear the transformation needs to be strongly nonlinear
514
(table 11.3)		 MULT: Performs the fixed-point multiplication of arg1 by arg2 and gives a 16-bit scaled (shifted left by 15) result. MULT: Performs the fixed-point multiplication of arg1 by arg2 and gives a 16-bit scaled (shifted right by 15) result.
535 Pennabaker and Mitchell [1993] Pennabaker and Mitchell [1992]
536 Telkap [1995] Tekalp [1995]
536 paper by Gall [1991] paper by Le Gall [1991]
537 presentation by Mangione-Smith and Memik [2002] presentation by Mangione-Smith and Memik [2001]
537 Cidecyan et al. [1992] Cideciyan et al. [1992]
537 Lyi and Tomizuka [1999] Yi and Tomizuka [1999]
545 the public-domain SoftFloat package the freely available SoftFloat package
545 See http://www.cs.berkeley.edu/~jhauser See http:/www.jhauser.us/arithmetic/SoftFloat.html
548 (A.3) [ missing ] ANDL {t|b} = s1, {s2 | im}              Logical AND
551 c0  br  $b0.0, L1 c0  brf  $b0.0, L1
570 when using the K&R mode (-KR switch) when using the K&R mode (-K+R switch)
577 #pragma if_prob(20,100) #pragma if_prob(20,100) // predicts "then" (foo) taken 20%
578 Assert less-than, ..., of the two expressions, exp1 and exp2, modulo n, in the current scope Assert less-than, ..., of the two expressions, exp1 and exp2, in the current scope
580 (i.e. the ivdep was asserted incorrectly) (i.e. the restrict keyword was used incorrectly)
582 #pragma stream
typedef int *stream_int_ptr;
stream_int_ptr *p;		 #pragma stream
typedef int *stream_int_ptr;
stream_int_ptr p;
582 #pragma stream
typedef int stream_int;
#pragma stream
typedef stream_int *stream_int_stream_ptr;
stream_int_stream_ptr *p;		 #pragma stream
typedef int stream_int;
#pragma stream
typedef stream_int *stream_int_stream_ptr;
stream_int_stream_ptr p;
592 (no induction variable dependences) (ignore vector dependences)
594 Profile-based optimization (...) is a two-step compilation process Profile-based optimization (...)  is a three-step compilation process
600
(Table A.12)		 lg2lineSize             ... (e.g. 8==> 64KB)
lg2StrLineSize        ... (e.g. 8==> 64KB)
lg2ICacheLineSize  ... (e.g. 8==> 64KB) 		lg2lineSize             ... (e.g. 8==> 64B)
lg2StrLineSize        ... (e.g. 8==> 64B)
lg2ICacheLineSize  ... (e.g. 8==> 64B)
605 # Compile modules with "risc" machine model
/home/vex/bin/cc -ms -O3 -fmm=verywide.mm -c file1.c
/home/vex/bin/cc -ms -O3 -fmm=verywide.mm -c file2.c 		# Compile modules with "verywide" machine model
/home/vex/bin/cc -ms -O3 -fmm=verywide.mm -width 4 -c file1.c
/home/vex/bin/cc -ms -O3 -fmm=verywide.mm -width 4 -c file2.c
615 (loops completed per cycle) (loop iterations completed per cycle)
617 pipes and name pipes pipes and named pipes
653 [missing Sander [1995] reference, cited at page 558] Sander (1995). G. Sander. VCG -- Visualisation of Compiler Graphs. Technical Report A01-95 Universität des Saarlandes, FB 14 Informatik, Germany, Feb. 1995 (available at http://rw4.cs.uni-sb.de/~sander/)

Submit new errata

Let us know about errors by writing to:   

[ Home ] [ Meet the Authors ] [ Errata Corrige ] [ Suggestions ] [ Table of Contents ] [ Downloads ] [ VEX forum ]